The Treatment of Acute Bronchitis With Trimethoprim and Sulfamethoxazole

Peter Franks, MD, and J. Arthur Gleiner, MDRochester, New York

Sixty-seven previously healthy patients with acute bronchitis were randomized and treated with either a fixed dose of tri­methoprim and sulfamethoxazole or placebo for seven days. All outcomes examined showed a trend favoring the use of an­tibiotic, with statistically significant differences for cough, night cough, mean temperature, and use of antihistamines or decongestants. Night cough occurred on 84 percent of nights in the control group vs 56 percent in the antibiotic group (P= -003). Cough occurred on 99 percent of days for patients in the control group vs 93 percent of days for patients in the antibiotic group (P=.05). Mean temperature over the seven nights was 37.3° C in the control group vs 36.9° C in the antibi­otic group (P=.004). The use of antihistamines and decon­gestants was reduced from 32 percent of days in the control group to 6 percent of days in the antibiotic group (P = .005). Patients in the antibiotic group worked 73 percent of days vs 55 percent of days for patients in the control group, which was significant when patients were stratified by the appearance of their sputum on Gram stain (P= .006). Smoking history was not found to help predict the response to antibiotic therapy.

Acute bronchitis in previously healthy adults is a frequently encountered problem in general med­ical practice. Acute bronchitis is usually defined clinically, based on history and physical examina­tion, as acute cough with sputum in the absence of pneumonia. Antibiotics are often prescribed empirically, usually without laboratory confirma­tion of pathogenesis. Despite the wealth of data demonstrating the usefulness of antibiotics in treatment of acute exacerbations of chronic bron­chitis, there are few data to support the routine use of antibiotics in acute bronchitis. Most studies that have purported to demonstrate the effect of anti-

From the Department of Preventive, Family, and Rehabilita­tion Medicine, and the Department of Medicine, University of Rochester School of Medicine and Dentistry, Rochester, New York. Requests for reprints should be addressed to Dr. Peter Franks, The Family Medicine Program, 885 South Avenue, Rochester, NY 14620.

biotics have compared two or more antibiotics without placebo control.110 Data from these stud­ies do not clearly support preference for any single agent and shed no light on whether there is any benefit of antibiotic compared with placebo. The only placebo-controlled study11 showed doxycy- cline to be no more effective than placebo in re­ducing the duration of cough, sputum production, or time lost from work. Since antibiotic therapy carries actual and potential costs to the patient, it is essential to demonstrate that the routine use of antibiotics for this illness is justified.

Several antibiotics are reasonable choices for treatment of acute bronchitis based on their in vitro activities and evidence of clinical safety.1215 The combination of trimethoprim and sulfameth­oxazole was studied because of theoretical advan­tages over other commonly used antibiotics.15'18 These advantages include excellent absorption, achievement of tissue and sputum levels equal to

® 1984 Appleton-Century-Crofts

THE JOURNAL OF FAMILY PRACTICE, VOL. 19, NO. 2: 185-190, 1984 185

TREA TMENT OF BRONCHITIS

or greater than serum levels, and peak sputum levels in excess of the minimal inhibitory concen­trations for common bacterial pathogens impli­cated in acute bronchitis. In addition, trimetho­prim and sulfamethoxazole are relatively safe and well tolerated and have a convenient dosage schedule. A double-blind placebo-controlled study of the effectiveness of using trimethoprim and sulfamethoxazole to treat the clinical syndrome of acute bronchitis was, therefore, undertaken.

The major hypothesis was that patients treated with trimethoprim and sulfamethoxazole would have a shorter duration of illness as measured by cough, sputum production, fever, and general sense of well-being. Because it was anticipated that the admission criteria would not differentiate between subjects with bacterial bronchitis and those with viral etiologies, a second hypothesis was that performing sputum Gram stains would allow a more accurate prediction of subjects who would benefit from antibiotic therapy.

MethodsThis study was conducted at the Family Medi­

cine Center of Highland Hospital, a family prac­tice with a patient profile previously found to be representative of the patient population of Roches­ter, New York.19 All outpatients who came in for appointments with their usual provider were eligi­ble for the study if they were aged 14 years or older and had a cough productive of sputum for less than 15 days. Patients were excluded from the study if they were pregnant, had a history of sulfa allergy, congestive heart failure, renal failure, or chronic pulmonary disease, or had had any sys­temic antibiotic within the two weeks prior to en­rollment. Patients were also excluded for clinical evidence of pneumonitis or inability to produce a sputum specimen at the time of enrollment. The diagnosis of acute bronchitis therefore rested on clinical grounds, ie, the presence of cough productive of sputum for less than 15 days in the absence of pneumonitis. To keep the methods closely parallel to the treatment of this type of patient in general practice, other laboratory tests were not performed. Informed consent was ob­tained from eligible patients after the nature of the study was explained, and then a sputum sample was obtained for Gram stain. Major baseline data

186

collected on patients are shown in Table 1. pa. tients were then randomly assigned to receive tri­methoprim and sulfamethoxazole (160/800 mg in a fixed-dose combination) twice daily for seven days or a placebo. The assignment to placebo or active group was done in a double-blind fashion, with the medications having been prepackaged, random­ized, and sequentially coded by the manufacturer. The tablets were identical in appearance. The code was broken at the time of data analysis or when the information was required for patient care. Pa­tients were enrolled in the study by their usual family physician, and at the time of enrollment the providers were asked to record any adjunctive therapy recommended to the patient. Providers were also asked to record whether they would have prescribed an antibiotic had the patient not been enrolled in the study.

Each patient was given a form on which to re­cord his or her symptoms during the seven-day course of the study drug. This form consisted of seven identical pages for the patient to make daily records of response to the study medication re­garding cough, sputum production, general well­being, fever, return to work, and use of adjunctive therapy. Items to be recorded each day were pres­ence of cough, presence of night cough, frequency of coughing spells (four or more spells per hour scored as 1, one to three spells per hour as 2, more than 10 spells per day but less than one per hour as 3, and fewer than 10 spells per day as 4); amount of cough (more than on day of office visit scored as 1, same as on day of office visit as 2, and less than on day of visit as 3); and temperature, activity level (in bed all day scored as 1, 25 percent of usual level as 2, 50 percent of usual level as 3, 75 percent of usual level as 4, and 100 percent of usual level as 5); for those employed, whether the patient had returned to work; use of adjunctive therapy; and the occurrence of side effects or other new symp­toms. Patients were asked to complete all items for seven consecutive days, even if all symptoms re­solved or if they had to drop out of the study for some other reason.

A Gram stain of sputum produced at the time of enrollment was examined for each patient and graded on the basis of organisms, polymorphonu­clear leukocytes seen, and evidence of contamina­tion. A Gram stain was considered to be “p°s*‘ tive” for bacterial infection if more than 20 polymorphonuclear leukocytes per oil field in

THE JOURNAL OF FAMILY PRACTICE, VOL. 19, NO. 2, 1984

TREATMENT OF BRONCHITIS

Table 1. Baseline Characteristics of Enrolled Patients

VariableActive Drug

(n = 34)Placebo (n = 33)

Age (mean years) 42 36

HistoryDuration of sym ptom s (mean days) 6 7Sputum yellow/green 87% 85%Smoking 43% 60%Fever (mean °F) 101.9 101.1Sore throat 75% 68%Coryza 79% 57%Dyspnea 26% 46%Chest pain 54% 42%Unable to work 45% 62%

ExaminationTemperature (mean °C) 37.1 37.25Pulse (mean beats/min) 80 73Pharyngitis 68% 44%Wheeze/rhonchi 13% 16%Gram stain positive (see text) 37.5% (n = 32) 30% (n = 30)

W ould have treated? 50% 48%

None of the differences were statistically significant (f tests fo r means, chi-square fo r all others). Gram stains on 5 patients were lost.

at least three fields were seen and if bacteria of no more than two morphological types were seen in the same fields as polymorphonuclear leu­kocytes. A Gram stain was considered to be con­taminated or indeterminate if epithelial cells were also seen in more than three fields. A Gram stain was read as “ negative” for bacterial infection in all other instances. All slides were stained, read, and reviewed by one of the authors (JAG).

The data were coded and analyzed by the au­thors using the SAS computer package.20 Baseline data were analyzed comparing drug and placebo groups and dropouts from both groups using chi- square tests, Fisher’s exact test, and / tests as appropriate. For each item in the diary a mean score over the seven days was devised, and the mean scores for the drug and placebo groups were compared using t tests. In addition, analysis of variance was used to determine whether there was an interaction between drug effect and history of smoking or between drug effect and Gram stain. Patients who did not complete all seven days of the symptom questionnaire were considered to

THE JOURNAL OF FAMILY PRACTICE, VOL. 19, NO. 2, 1984

have remained at the same level of function for each category for the remainder of the seven days as they were on the last day that they completed the form. Multivariate analysis of variance was used to analyze all dependent variables simultane­ously, both including dropouts as described above and excluding them.

ResultsA total of 67 patients were enrolled. There were

no significant differences between the antibiotic and placebo groups at the time of enrollment (Table 1).

Thirteen patients did not return their symptom forms—4 in the placebo group and 9 in the antibi­otic group (see Appendix for detailed information on these patients). Patients who did not return their forms were significantly more likely to have a positive Gram stain (P= .02, Fisher’s exact test). Of the 4 patients taking placebo who dropped out of the study, 3 had a positive Gram stain and 1

187

TREATMENT OF BRONCHITIS

Table 2. Mean Outcomes Over Seven Days of Study

VariableActive Drug

(n*)Placebo

(n*) p*«

Cough 93% (25) 99% (29) .05Night cough 56% (25) 84% (29) .003Cough frequencyt 2.6 (24) 2.35 (29) .18Cough a m o u n tt 2.4 (24) 2.2 (28) .19Tem perature (°C) 36.9 (23) 37.3 (22) .004A ctiv ity leve lt 3.6 (24) 3.4 (29) .29Return to w ork 74% (19) 60% (29) .09Expectorant/antitussive 32% (25) 42% (29) .17Analgesic/anti pyretic 40% (25) 48% (29) .27Antihistam ine/decongestant 6% (25) 29% (29) .005

*n = Num ber o f patients responding **t Test, one-tail p robability tSee text fo r explanation of scoring

patient’s Gram stain was lost. Of the 9 patients who dropped out of the antibiotic group, 5 had positive Gram stains, 3 had negative Gram stains, and 1 patient’s Gram stain was lost. There were no other significant differences between those who returned their forms and those who did not in any of the variables recorded at the time of enrollment. Four patients discontinued their medication before the end of seven days because of side effects (headaches and nausea), 1 in the placebo group and 3 on antibiotics. Two patients discontinued their medication because they felt completely well (1 in the placebo and 1 in the antibiotic group). Two patients withdrew from the study and stopped medication because of increased symptoms, 1 in the antibiotic and 1 in the placebo group.

Of the seven possible days for patients to have continued coughing, a mean of 99 percent of the placebo group and 93 percent of the antibiotic group recorded a cough (P = .05, t test, one tail). Of the remaining ten variables, all showed a trend favoring the antibiotic group. For three of these variables, the difference between antibiotic and placebo groups was statistically significant (Table 2). Patients on antibiotics showed a statistically significant difference in presence of night cough, fever, and use of antihistamine or decongestant adjunctive therapy. Analysis of variance did not show any benefit for subgroups stratified by his­tory of smoking. The use of analysis of variance to

188

stratify for the Gram stain findings showed, how­ever, that patients in the antibiotic group were significantly more likely to have returned to work by the end of the study (Table 3); that is, although overall the antibiotic did not show a statistically significant benefit, when the results were stratified by Gram stain, a statistically significant benefit emerged. No other result was affected by the re­sults of the Gram stain. The multivariate test of all dependent variables for the hypothesis of no over­all drug effect revealed a statistically significant benefit (P = .008, two-tailed test). When the analy­sis was repeated excluding all patients with incom­plete forms, the statistical outcome was unchanged.

DiscussionThis study was designed to evaluate the effec­

tiveness of trimethoprim and sulfamethoxazole in acute bronchitis diagnosed by history and physical examination. Other studies that have evaluated this problem have been inconclusive. In particu­lar, prior studies have been flawed by use of anti­biotics that are bacteriostatic rather than bacterici­dal and by lack of demonstration that the agent used achieved adequate levels in tissue and sputum. No attempt was made in prior studies to perform multivariate analysis. Although the posi­tive results in this study may have been biased by

THE JOURNAL OF FAMILY PRACTICE, VOL. 19, NO. 2, 1984

TREATMENT OF BRONCHITIS

Table 3. Mean Number (Percentage) of Days Worked

PlaceboNo. (%)

AntibioticNo. (%)

Gram StainPositive 6(36) 6(55)Negative 9(68) 11 (84)

Analysis o f VarianceDrug effect P = .006,1-tailGram stain effect P = .013,1-tailInteraction between drug P = .23, 2-tail

and Gram stain effect

the inclusion of patients with syndromes known to be responsive to antibiotics, such as pneumonia, the intent of the study was to investigate a syn­drome, usually defined clinically. Thus the results may be generalized to the clinical situation where the same bias occurs; that is, further tests such as chest x-ray films are not routinely obtained.

In the final analysis, an assessment of the use­fulness of trimethoprim and sulfamethoxazole in acute bronchitis must take into account many fac­tors with relatively soft end points. A strict cost- benefit analysis is useful, but provides only limited information to the clinician faced with the imme­diate problem of an ill patient in his office who is insisting that an antibiotic be prescribed. These data provide support for the use of trimethoprim and sulfamethoxazole in the treatment of patients with acute bronchitis because of the clinically im­portant and statistically significant benefits that were associated with the use of trimethoprim and sulfamethoxazole in patients with acute cough who produced sputum in the office. This benefit appears to be independent of prior smoking his­tory or of Gram stain findings in the sputum. It was hypothesized that a more accurate prediction could be made on the outcome of trimethoprim and sulfamethoxazole therapy by looking at the sputum Gram stains. There are several possible reasons that outcome appeared to be relatively in­dependent of Gram stain. Gram stains are subject to possible method errors such as inadequate or improper collection of specimens, sampling errors in preparing or viewing the slides, and inappropri­ate criteria for judgment of the results of the slide. No evidence of inconsistent reading of the slides

THE JOURNAL OF FAMILY PRACTICE, VOL. 19, NO. 2, 1984

was found. It remains possible that specimens did not reflect the clinical status of the patient, however, and in the office setting there does not appear to be a reliable way to circumvent this problem.

In view of the earlier return to work of patients who were treated with trimethoprim and sulfa­methoxazole, it is probable that treatment is cost effective considering the low cost and morbidity involved in therapy. However, no formal cost analysis was performed. Since acute bronchitis is a self-limited condition with a favorable outcome in the overwhelming majority of patients, and since there is some hazard to using antibiotics, this issue warrants further investigation.

The problem of statistically analyzing multiple outcomes is complex and unresolved. The use of multiple outcomes is desirable to allow adequate evaluation of response in situations where all of the end points are relatively “ soft.” There are multiple ways to aggregate data to simplify the presentation. Rather than developing a scoring system to combine all results into a synthesized outcome, it was decided to present the results as shown in Table 2 to allow the reader to fashion his or her own interpretation. The probability of four results significant at a level of P < .05 in a series of 29 independent tests is P = .05.21 The probability of 11 independent tests all having a trend in the same direction is P < .001. The tests reported here, however, are not independent, since some of the outcomes are correlated. The multivariate test re­sult that takes account of the correlation between the outcome variables is compatible with these re­sults. The multivariate test assigns equal weight to each outcome variable that is undesirable on clini­cal grounds.

These results differ from those of Stott and West,11 who found no benefit from the use of doxycycline in the treatment of patients similar to those reported here. The differences found may reflect antibiotics used, difference in patient popu­lations (since patients in that study were not required to produce sputum prior to enrollment), or inadequate discriminatory power in study design, since a nonsignificant trend in favor of doxycy­cline was found in that study.

Although these data are incomplete, no evi­dence that a bias was introduced to explain the results was found. To explain the results on the basis of systematic bias, one would have to hy-

189

TREATMENT OF BRONCHITIS

pothesize either that patients taking trimethoprim and sulfamethoxazole who dropped out did much worse than those who remained in the study, or that those who took placebo and dropped out fared much better. These hypotheses seem equally un­likely in that although patients who dropped out were more likely to have positive Gram stains, there was no evidence that within the group who dropped out those who were taking trimethoprim and sulfamethoxazole were significantly more ill as judged by any of the baseline parameters.

The results of this study support the use of tri­methoprim and sulfamethoxazole in patients with cough and sputum production who have no clinical evidence of pneumonitis and no contraindications to this medication. The routine use of Gram stain in this group of patients is not recommended, since the use of this test does not improve the clinician's ability to predict response to antibiotic therapy with trimethoprim and sulfamethoxazole.

AcknowledgmentThe trimethoprim and sulfamethoxazole used in this

study were provided by Hoffmann-La Roche, Inc.

References1. Klimek JJ: Use of oral cephalosporins in the treat­

ment of upper and lower respiratory tract infections: An overview. J Int Med Res 8(suppl 1):77, 1980

2. Mogabgab WJ, Pollock B, Beville RB, et al: Treat­ment of acute bacterial bronchitis and pneumonia with cefaclor. Postgrad Med J 55(suppl 4):62, 1979

3. Brodie NH, McGhie RL, O'Hara H, et al: A double­blind study comparing cefaclor and amoxycillin in the treatment of respiratory tract infections in general practice. Pharmatherapeutica 2:494, 1980

4. Jaffe GV, Grimshaw JJ: Comparative study of cefa­clor and amoxycillin in the treatment of respiratory tract

infections in general practice. Curr Med Res ODin fi-tm 1980 °S'

5. Pines A, Raafat H: A comparative clinical trial oftetroxoprim/sulphadiazine and ampicillin in the treatment of purulent bronchial infections. J Antimicrob ChemnttJr 5(suppl B):201, 1979 er

6. Lee GS, Kakati R, Mahmoud IA: A comparison of cefaclor and ampicillin in the treatment of respiratory in­fection in elderly in-patients. Curr Med Res Onin 6-RM 1980

7. Ferber H, Ahrens KH: A short-term study of tetroxo­prim/sulphadiazine in the treatment of acute bronchitis and urinary tract infections. J Antimicrob Chemother 5(suppl B):231, 1979

8. Cooper J, McGillion FB, West B: A comparison of cotrimoxazole and amoxycillin in acute bronchitis Practi tioner 220:798, 1978

9. Murphy JE, Donald JF, Molla AL: A single-blind comparative clinical trial of lymecycline and amoxycillin in the treatment of acute bronchitis in general practice J Int Med Res 4:64, 1976

10. Carroll PG, Krejci SP, Mitchell J, et al: A compara­tive study of co-trimoxazole and amoxycillin in the treat­ment of acute bronchitis in general practice. Med J Aust 2:286, 1977

11. Stott NC, West RR: Randomized controlled trial of antibiotics in patients with cough and purulent sputum Br Med J 2:556, 1976

12. Saginur R, Bartlett JG: Antimicrobial drug suscep­tib ility of respiratory isolates of Hemophilus influenzae from adults. Am Rev Resp Dis 122:61, 1980

13. Ingold A: Sputum and serum levels of amoxycillin in chronic bronchial infections. Br J Dis Chest 69:211, 1973

14. Levin S, Trenholme GM: Use of antibiotics in the outpatient setting. Primary Care 6:3, 1979

15. Wong GA, Pierce TH, Goldstein E, Hoeprich PD: Penetration of antimicrobial agents into bronchial secre­tions. Am J Med 59:219, 1975

16. Bach MC, Finland M, Gold O, Wilcox C: Susceptibil­ity of recently isolated pathogenic bacteria to trimethoprim and sulfamethoxazole separately and combined. J Infect Dis 128(suppl):S508, 1973

17. Bushby SRM: Trimethoprim-sulfamethoxazole: In vitro microbiological aspects. J Infect Dis 128(suppl):S442, 1973

18. Hansen I, Lykkegaard N, Bertelsen S: Trimethoprim in human saliva, bronchial secretion and lung tissue. Acta Pharmacol Toxicol 32:337, 1973

19. Froom J, Culpepper L, Kirkwood CR, et al: An inte­grated medical record and data system for primary care: Part 4: Family information. J Fam Pract 5:265, 1977

20. Helwig JT, Council KA (eds): SAS User's Guide: 1979 edition. Raleigh, NC, SAS Institute, 1979

21. Rumke CL: How long is the wait for an uncommon event? J Chronic Dis 35:561, 1982

Appendix

Patients who did not return their diaries are reported. They were followed up by chart review, telephone, and letter.

Trimethoprim and Sulfamethoxazole Group: Two patients had their codes broken within 24 hours—one developed nausea and vomiting with the medication as well as streptococcal pharyngitis requiring penicillin therapy; the other developed otitis media and was continued on trimethoprim-sulfamethoxazole, on which his symptoms resolved; two subsequently reported that they “ did OK" but did not follow up; one reported that the medication did not help but did not follow up; no follow-up was available on the remaining four.

Placebo Group: One patient reported getting worse and received care elsewhere, one patient reported not taking the medication because of fear of the side effects, and no follow-up was available on the remaining two.

190 THE JOURNAL OF FAMILY PRACTICE, VOL. 19, NO. 2, 1984